Skip to contentSkip to conversation

externalsrc: git submodule--helper list unsupported

PATCH New Peter Kjellerstedt 1 participant 3 messages Nov 5, 2022

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 externalsrc: git submodule--helper list unsupported

From: John Edward Broadbent <[email protected]>

Git has removed support for "git submodule--helper list".
https://github.com/git/git/commit/31955475d1c283120d5d84247eb3fd55d9f5fdd9

This change provides an alternate method for gathering the submodules information.

Tested:
Build recipes with and without submodules

---
 meta/classes/externalsrc.bbclass | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)
meta/classes/externalsrc.bbclass +10 -9
index 8136d25cb1..766c4979c5 100644
@@ -229,15 +229,16 @@ def srctree_hash_files(d, srcdir=None):
             env['GIT_INDEX_FILE'] = tmp_index.name
             subprocess.check_output(['git', 'add', '-A', '.'], cwd=s_dir, env=env)
             git_sha1 = subprocess.check_output(['git', 'write-tree'], cwd=s_dir, env=env).decode("utf-8")
-            submodule_helper = subprocess.check_output(['git', 'submodule--helper', 'list'], cwd=s_dir, env=env).decode("utf-8")
-            for line in submodule_helper.splitlines():
-                module_dir = os.path.join(s_dir, line.rsplit(maxsplit=1)[1])
-                if os.path.isdir(module_dir):
-                    proc = subprocess.Popen(['git', 'add', '-A', '.'], cwd=module_dir, env=env, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
-                    proc.communicate()
-                    proc = subprocess.Popen(['git', 'write-tree'], cwd=module_dir, env=env, stdout=subprocess.PIPE, stderr=subprocess.DEVNULL)
-                    stdout, _ = proc.communicate()
-                    git_sha1 += stdout.decode("utf-8")
+            if os.path.exists(".gitmodules"):
+                submodule_helper = subprocess.check_output(["git", "config", "--file", ".gitmodules", "--get-regexp", "path"], cwd=s_dir, env=env).decode("utf-8")
+                for line in submodule_helper.splitlines():
+                    module_dir = os.path.join(s_dir, line.rsplit(maxsplit=1)[1])
+                    if os.path.isdir(module_dir):
+                        proc = subprocess.Popen(['git', 'add', '-A', '.'], cwd=module_dir, env=env, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+                        proc.communicate()
+                        proc = subprocess.Popen(['git', 'write-tree'], cwd=module_dir, env=env, stdout=subprocess.PIPE, stderr=subprocess.DEVNULL)
+                        stdout, _ = proc.communicate()
+                        git_sha1 += stdout.decode("utf-8")
             sha1 = hashlib.sha1(git_sha1.encode("utf-8")).hexdigest()
         with open(oe_hash_file, 'w') as fobj:
             fobj.write(sha1)
#1 lore raw

PATCH 2/3 externalsrc.bbclass: fix git repo detection

From: Martin Jansa <[email protected]>

* fix issue introduced in:
https://git.openembedded.org/openembedded-core/commit/?id=95fbac8dcad6c93f4c9737e9fe13e92ab6befa09

* it added check for s_dir + git-dir (typically '.git') isn't the same as ${TOPDIR} + git-dir, but due to copy-paste issue it was just comparing it with s_dir + git-dir again, resulting in most external repos (where git-dir is '.git') to be processed as regular directory (not taking advantage of git write-tree).

* normally this wouldn't be an issue, but for big repo with a lot of files this added a lot of checksums in:
d.setVarFlag('do_compile', 'file-checksums', '${@srctree_hash_files(d)}')

and I mean *a lot, e.g. in chromium build it was 380227 paths which still wouldn't that bad, but the checksum processing in siggen.py isn't trivial and just looping through all these checksums takes very long time (over 1000sec on fast NVME drive with warm cache) and then
https://git.openembedded.org/bitbake/commit/?id=b4975d2ecf615ac4c240808fbc5a3f879a93846b
made the processing a bit more complicated and the loop in get_taskhash() function took 6448sec and to make things worse there was no output from bitbake during that time, so even with -DDD it looks like this:

DEBUG: virtual/libgles2 resolved to: mesa (langdale/oe-core/meta/recipes-graphics/mesa/mesa_22.2.0.bb) Bitbake still alive (no events for 600s). Active tasks:
Bitbake still alive (no events for 1200s). Active tasks:
Bitbake still alive (no events for 1800s). Active tasks:
Bitbake still alive (no events for 2400s). Active tasks:
Bitbake still alive (no events for 3000s). Active tasks:
Bitbake still alive (no events for 3600s). Active tasks:
Bitbake still alive (no events for 4200s). Active tasks:
Bitbake still alive (no events for 4800s). Active tasks:
Bitbake still alive (no events for 5400s). Active tasks:
Bitbake still alive (no events for 6000s). Active tasks:
DEBUG: Starting bitbake-worker

without -DDD it will get stuck for almost 2 hours in:
"Initialising tasks..."
before it finally writes sstate summary like:
"Sstate summary: Wanted 3102 Local 0 Mirrors 0 Missed 3102 Current 1483 (0% match, 32% complete)"

* fix the copy&paste typo to use git work-tree in most cases, but be aware that this issue still exists for huge local source trees not in git

[YOCTO #14942]

---
 meta/classes/externalsrc.bbclass | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
meta/classes/externalsrc.bbclass +2 -2
index 766c4979c5..686553f900 100644
@@ -211,8 +211,8 @@ def srctree_hash_files(d, srcdir=None):
     try:
         git_dir = os.path.join(s_dir,
             subprocess.check_output(['git', '-C', s_dir, 'rev-parse', '--git-dir'], stderr=subprocess.DEVNULL).decode("utf-8").rstrip())
-        top_git_dir = os.path.join(s_dir, subprocess.check_output(['git', '-C', d.getVar("TOPDIR"), 'rev-parse', '--git-dir'],
-            stderr=subprocess.DEVNULL).decode("utf-8").rstrip())
+        top_git_dir = os.path.join(d.getVar("TOPDIR"),
+            subprocess.check_output(['git', '-C', d.getVar("TOPDIR"), 'rev-parse', '--git-dir'], stderr=subprocess.DEVNULL).decode("utf-8").rstrip())
         if git_dir == top_git_dir:
             git_dir = None
     except subprocess.CalledProcessError:
#2 lore raw

PATCH 3/3 externalsrc.bbclass: Remove a trailing slash from ${B}

In reply to Peter Kjellerstedt

From: Peter Kjellerstedt <[email protected]>

The trailing slash in ${B} caused -fdebug-prefix-map=${B}=... to not match as intended, resulting in ${TMPDIR} ending up in files in ${PN}-dbg when externalsrc was in use, which in turn triggered buildpath QA warnings.

---
 meta/classes/externalsrc.bbclass | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/classes/externalsrc.bbclass +1 -1
index 686553f900..a6a8ca6318 100644
@@ -60,7 +60,7 @@ python () {
         if externalsrcbuild:
             d.setVar('B', externalsrcbuild)
         else:
-            d.setVar('B', '${WORKDIR}/${BPN}-${PV}/')
+            d.setVar('B', '${WORKDIR}/${BPN}-${PV}')
 
         local_srcuri = []
         fetch = bb.fetch2.Fetch((d.getVar('SRC_URI') or '').split(), d)